Predicting School Dropout Using Machine Learning Models: A Case Study of Nyanza District, Rwanda

Abstract

This study investigates the application of machine learning models to predict school dropouts in Nyanza District, Rwanda, addressing the challenge of early identification of at-risk students. By adopting a classification-based approach, the research analyzes data from parents  or guardians, school instructors, and teachers to pinpoint contributing factors such as socioeconomic conditions, academic performance,  and family background. The research explores a range of machine learning models, including Logistic Regression, Decision Tree  Classifier, Gradient Boosting Regression, Artificial Neural Networks (ANN), K-Nearest Neighbors (KNN), and Naive Bayes. These models  are evaluated using metrics like accuracy, recall, precision, F1 score, and ROC-AUC, with an emphasis on balancing recall (identifying at- risk students) and precision. The study reveals that different models offer varying levels of performance. KNN achieves a notable  accuracy of 0.72 and an exceptional recall of 0.91, successfully identifying 91% of at-risk students. Naive Bayes, however, is highlighted as  the most well rounded model, balancing precision and recall effectively. This research fills the gap in predictive analytics for dropout  prevention in Nyanza District and offers actionable insights for educators and policymakers to enhance student retention through  targeted interventions.